The present invention relates generally to a computer implemented method, data processing system, and computer program product for monitoring, measurement, and management of the efficiency of electrical power delivery systems. More specifically, the present invention relates to detecting degradation in power supplies in a data center and proactively removing degraded power supplies from service or otherwise modifying power delivery or system loads to enhance productive use of electricity.
There are over 1.5 billion power supplies used to convert alternating current (AC) into direct current (DC) useful for devices such as televisions, cellular phones, and computers. Approximately 11% of electricity in the U.S. flows through power supplies. Most power supplies are between 20-90% efficient in converting AC to DC. In addition to wasting electricity, an inefficient power supply can produce unwanted heat. In a data center with a high density of information technology (IT) equipment, removing this extra heat results in additional cost on top of the increased IT electricity consumption due to power conversion inefficiency.
Many factors may impact efficiency of power supplies. Contamination and component wear can reduce efficiency during the life of a power supply. In addition, dust and humidity may clog a power supply's cooling fans. Dust collected on the power supply's heat sink surface reduces its efficiency in removing heat. This gradual deterioration in the power supply's fan and heat sink performance results in raising the temperature of the power supply, which leads to reducing the power supply efficiency. In addition, surges and voltage variations in the AC arriving from the mains (i.e. the chief power lines entering a data center) also limit the efficiency of a power supply. All these reasons add uncertainty to the actual run-time power supply efficiency. Eventually, degradation of a power supply can lead to its failure and disruption of the IT equipment operation. Without monitoring the actual power supply efficiency at run time, there is no way to detect its efficiency deterioration, predict failure, or, in some cases, even discover its failure.
Accordingly, a remedy to the situation is warranted.